<p><span style="font-size:15px"><span>本文分享一篇知識圖譜表示學(xué)習(xí)匯報</span><span>ppt</span><span>，將知識圖譜表示學(xué)習(xí)方法粗略分為四大類考传，涉及將近</span><span>30</span><span>篇優(yōu)秀論文盲厌，只簡單介紹其核心思想坚俗，完整匯報</span><span>ppt</span><span>獲取請關(guān)注公眾號【AI機器學(xué)習(xí)與知識圖譜】回復(fù)關(guān)鍵字：</span></span><span style="font-size:15px"><strong>知識圖譜表示學(xué)習(xí)</strong></span></p><p>
</p><p><strong><span style="font-size:15px"><span>1</span><span>、</span><span>翻譯距離模型</span></span></strong><span style="font-size:15px"><span>：包括</span><span>TransH</span><span>尸昧、</span><span>TransR</span><span>揩页、</span><span>TransD</span><span>、</span><span>TranSparse</span><span>烹俗、</span><span>TransM</span><span>爆侣、</span><span>MianfoldE</span><span>、</span><span>TransF</span><span>幢妄、</span><span>TransA</span><span>兔仰、</span><span>KG2E</span><span>、</span><span>TransG</span><span>磁浇、</span><span>UM</span><span>斋陪、</span><span>SE</span><span>模型等；</span></span></p><p><strong><span style="font-size:15px"><span>2</span><span>置吓、語義匹配模型</span></span></strong><span style="font-size:15px"><span>：包括</span><span>RESCAL</span><span>无虚、</span><span>DistMult</span><span>、</span><span>HoLE</span><span>衍锚、</span><span>ComplEx</span><span>友题、</span><span>ANALOGY</span><span>、</span><span>SNE</span><span>戴质、</span><span>NTN</span><span>度宦、</span><span>MLP</span><span>、</span><span>NAM</span><span>模型等告匠；</span></span></p><p><strong><span style="font-size:15px"><span>3</span><span>戈抄、隨機游走模型</span></span></strong><span style="font-size:15px"><span>：包括</span><span>DeepWalk</span><span>、</span><span>LINE</span><span>后专、</span><span>node2vec</span><span>模型等划鸽；</span></span></p><p><strong><span style="font-size:15px"><span>4</span><span>、子圖匯聚模型</span></span></strong><span style="font-size:15px"><span>：包括</span><span>GCN</span><span>、</span><span>GAT</span><span>裸诽、</span><span>GraphSage</span><span>模型等嫂用。</span></span></p><p><span style="font-size:15px">
</span></p><p><span style="font-size:20px"><strong>Motivation</strong></span></p><p>
</p><p><span style="font-size:15px"><span>知識圖譜是由實體（節(jié)點）和關(guān)系（不同類型的邊）組成的多關(guān)系圖，每條邊連接頭尾兩個實體丈冬，通常用</span><span>SPO</span><span>三元組進行表示（</span><span>subject,predicate, object</span><span>）嘱函，被稱為一個事實。雖然知識圖譜在表示結(jié)構(gòu)化數(shù)據(jù)方面很有效埂蕊，但這類三元組的潛在符號特性通常使得</span><span>KGs</span><span>很難操作往弓。</span></span></p><p>
</p><p><span style="font-size:15px"><span>因此知識圖譜表示學(xué)習(xí)便成為了一個熱門的研究方向，知識圖譜嵌入的關(guān)鍵思想是將圖譜中的實體</span><span>entity</span><span>和關(guān)系</span><span>relation</span><span>轉(zhuǎn)化為連續(xù)的向量粒梦，在保留</span><span>KG</span><span>原有結(jié)構(gòu)的同時使得操作方便亮航。于是便可將</span><span>entityembedding</span><span>和</span><span>relationembedding</span><span>用到下游各種任務(wù)中，例如圖譜補全匀们，關(guān)系抽取缴淋，實體分類，實體鏈接及實體融合等</span></span></p><p>
</p><p><span style="font-size:15px">知識圖譜嵌入技術(shù)經(jīng)典三個步驟：</span></p><p><span style="font-size:15px"><span>1</span><span>泄朴、</span><span>representingentities and relations</span></span></p><p><span style="font-size:15px"><span>2</span><span>重抖、</span><span>defininga scoring function</span></span></p><p><span style="font-size:15px"><span>3</span><span>、</span><span>learningentity and relation representations</span><span>（最大化所有觀測事實的置信度</span><span>plausibility</span><span>）</span></span></p><p><span style="font-size:15px"><span>根據(jù)</span><span>scoringfunction</span><span>區(qū)別分為</span><span>distance-based scoring functions</span><span>和</span><span>similarity-based scoring functions</span></span></p><p><span style="font-size:15px">
</span></p><p><span style="font-size:20px"><strong>Papers</strong></span></p><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-7705c15ad6925e9a.jpeg" img-data="{"format":"jpeg","size":99231,"height":600,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-5e1bedab6303c7c7.jpeg" img-data="{"format":"jpeg","size":40440,"height":585,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-88995591f0a469d6.jpeg" img-data="{"format":"jpeg","size":95862,"height":579,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-89aa2cd4b9ac6794.jpeg" img-data="{"format":"jpeg","size":44012,"height":523,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-ef1052927f117779.jpeg" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-d73958dc3d41d0be.jpeg" img-data="{"format":"jpeg","size":86715,"height":641,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-43548c26a7e66bb6.jpeg" img-data="{"format":"jpeg","size":81019,"height":596,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-6ebeb7f7c7c8c8e5.jpeg" img-data="{"format":"jpeg","size":52391,"height":548,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-b334e44e5acf7d00.jpeg" img-data="{"format":"jpeg","size":87769,"height":586,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-2b800a1507cdc8a4.jpeg" img-data="{"format":"jpeg","size":104003,"height":579,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-9efe7e43893b7495.jpeg" img-data="{"format":"jpeg","size":57531,"height":544,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-5d9cc168cdb1a1eb.jpeg" img-data="{"format":"jpeg","size":84030,"height":610,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-c165455e981c0b15.jpeg" img-data="{"format":"jpeg","size":91587,"height":606,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-32bad0844fb20448.jpeg" img-data="{"format":"jpeg","size":104973,"height":570,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-f6ffe15483aadc7a.jpeg" img-data="{"format":"jpeg","size":71939,"height":587,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-4555a84da4508957.jpeg" img-data="{"format":"jpeg","size":35470,"height":321,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-53ec1a824148ec36.jpeg" img-data="{"format":"jpeg","size":61945,"height":589,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-fa69d7fc905c7b8b.jpeg" img-data="{"format":"jpeg","size":54526,"height":494,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-1dfc4c848e2f6adb.jpeg" img-data="{"format":"jpeg","size":62667,"height":465,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-b8e5233cc37be995.jpeg" img-data="{"format":"jpeg","size":66425,"height":554,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><div class="image-package"><img src="https://upload-images.jianshu.io/upload_images/26011021-23915df046d3981e.jpeg" img-data="{"format":"jpeg","size":84630,"height":594,"width":1080}" class="uploaded-img" style="min-height:200px;min-width:200px;" width="auto" height="auto"/>
</div><p>
</p><p>
</p><p><span style="font-size:18px"><strong>往期精彩</strong></span></p><p>
</p><p><span style="font-size:14px">【知識圖譜系列】探索DeepGNN中Over-Smoothing問題</span>
</p><p/><p><span style="font-size:14px">【知識圖譜系列】多關(guān)系神經(jīng)網(wǎng)絡(luò)CompGCN</span></p><p>各大AI研究院共35場NLP算法崗面經(jīng)奉上
</p><p/><p>干貨 | Attention注意力機制超全綜述</p><p><span style="font-size:14px"/></p><p><span style="font-size:14px">干貨|一文弄懂機器學(xué)習(xí)中偏差和方差</span></p><p><span style="font-size:14px">Transformer模型細(xì)節(jié)理解及Tensorflow實現(xiàn)</span>
</p><p><span style="font-size:14px"/></p><p><span style="font-size:14px">機器學(xué)習(xí)算法篇：最大似然估計證明最小二乘法合理性</span>
</p><p><span style="font-size:14px"/></p><p>
</p>